Forecast Model Comparisons of ISO New England Electricity Demand

• Activity Number: 227
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Timothy Fletcher*+
• Companies: Northwestern University
• Keywords: electricity demand forecasting ; ARIMA ; artificial neural networks ; support vector machines ; random forests

Abstract

Reliable access and constant availability to electricity is important for the economy and safety of the residents of New England. The Independent System Operators (ISO) New England is responsible for meeting the electricity needs of the 6.5 million households and businesses in the New England region by managing the wholesale electricity market and overseeing the operation of the power generation and transmission system. ISO New England (ISO-NE) is composed of eight zones. The objective of this paper is to compare methods of forecasting short-term electricity demand in ISO-NE. We are asking: Can a predictive model be constructed that improves the Day Ahead Demand (DAHD) forecast? ARIMA, Artificial Neural Networks, Support Vector Machines, and Random Forests modeling techniques are tested on a data set for the calendar year 2012. Each of these approaches is compared for their predictive ability. Each zone within ISO-NE is modeled individually. The RMSE of the demand forecasts for each model are compared to the DAHD RMSE during the same period. Using one or more of the modeling tools can improve the DAHD hourly forecast values.